This survey is to help Anuket-Thoth to Stay 'relevant' to Telco's Needs
Are you familiar with the AI/ML works (solution and systems) present in your organization?
Vote |
---|
changeableVotes | true |
---|
title | Familiarity with your Organization's AI/ML works |
---|
showComments | true |
---|
|
YES.
NO. |
MOTORS: MOdels - TOols - ResearchStudies
What is more valuable to Telcos?
Vote |
---|
changeableVotes | true |
---|
title | We (Thoth) build better (explainable, least-error, accurate) ML-models to important problems |
---|
showComments | true |
---|
|
5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
Vote |
---|
changeableVotes | true |
---|
title | We (Thoth) optimize the existing model for important problems |
---|
showComments | true |
---|
|
5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
Vote |
---|
changeableVotes | true |
---|
title | We (Thoth) build a flexible* ML-Framework with implementation of important models that can be trained and used. |
---|
showComments | true |
---|
|
5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
Vote |
---|
changeableVotes | true |
---|
title | We (Thoth) build tools that can help telcos to build ML-Models |
---|
showComments | true |
---|
|
5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
Vote |
---|
changeableVotes | true |
---|
title | We (Thoth) conduct research studies with thorough and systematic analysis of Open-Problems |
---|
showComments | true |
---|
|
5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
Dataset
If we generate dataset in testbeds (Openstack/Kubernetes + Chaos Tools such as litmus + OSS TrafficGens) and use it to train the ML-Models (ex: failure prediction, anomaly detection, etc), how relevant these models will be useful for production networks of Telcos?
Vote |
---|
changeableVotes | true |
---|
title | Relevance of Dataset + Model from Testbed (labs) to the Production |
---|
showComments | true |
---|
|
Yes. Relevant
No. It will not be.
Why do it with testbed dataset when we (Telco) have access to production dataset? |
Frameworks
Telcos have (for past 15-20 yrs - maybe even more) AI/ML systems solving business problems (customer churns) and/or legacy Network management (failures prediction, anomaly detection, etc.). Is it Important that the ML-models we develop (related to Telco Clouds) should work in these frameworks (of legacy systems) ?
Vote |
---|
changeableVotes | true |
---|
title | Integration with legacy AI/ML Systems? |
---|
showComments | true |
---|
|
YES (we don't want to maintain multiple systems)
NO (Managed by different silos) |
Models: Importance of AI/ML Problems related to NFV
...
Please add any other problems you consider that is important here:
|
---|
Vote |
---|
changeableVotes | true |
---|
title | Importance of Failure Prediction (VM) |
---|
showComments | true |
---|
| 5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
|
Vote |
---|
changeableVotes | true |
---|
title | Importance of Failure Prediction (Containers) |
---|
showComments | true |
---|
| 5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
|
Vote |
---|
changeableVotes | true |
---|
title | Importance of Failure Prediction (Links) |
---|
showComments | true |
---|
| 5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
|
Vote |
---|
changeableVotes | true |
---|
title | Importance of Failure Prediction (Apps) |
---|
showComments | true |
---|
| 5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
|
Vote |
---|
changeableVotes | true |
---|
title | Importance of Failure Prediction (Services - control/middleware/) |
---|
showComments | true |
---|
| 5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
|
Vote |
---|
changeableVotes | true |
---|
title | Importance of Traffic Engineering (Using Reinforcement LearningEx: Reinforcement Learning) |
---|
showComments | true |
---|
| 5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
|
Vote |
---|
changeableVotes | true |
---|
title | Importance of Other Predictions (resource utilization, attack, SLA-breach) |
---|
showComments | true |
---|
|
5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
|
Vote |
---|
changeableVotes | true |
---|
title | Importance of Traffic Engineering (Techniques other than Reinforcement LearningResource Optimization (scheduling, deployment scaling, migration, etc.) |
---|
showComments | true |
---|
| 5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
|
Vote |
---|
changeableVotes | true |
---|
title | Importance of Anomaly Detection (metrics, logs, stats) |
---|
showComments | true |
---|
| 5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
|
Vote |
---|
changeableVotes | true |
---|
title | Importance of Alert Filtering/classification - Finding Actionable Alerts |
---|
showComments | true |
---|
| 5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
|
Vote |
---|
changeableVotes | true |
---|
title | Importance of Correlation - as part of causal analysis |
---|
showComments | true |
---|
| 5. Very Important
4. Important
3. Good to have
2. Not important
1. No value addition. |
|
|
Panel |
---|
title | Comments and Suggestions |
---|
|
|